The paper batch sheet is not a clerical habit. It is a liability that sits on your P&L, your recall exposure, and your next audit. Here is the record that replaces it, field by field, and why moving it off paper is the first thing your data foundation needs.
Read this first · What this is for
This is a template you adopt, not a product pitch. The sections below describe what a complete batch record captures at the station, in the order the work moves through it. You can hand the full field list to your quality lead and build it in whatever system you already run.
There are no invented numbers on this page. The only external figure is a public rule with a link. The dollars in play are your own: your batch count, your giveaway per run, your labor hours re-keying sheets. We give you the structure. You put your numbers in.
Why the paper batch sheet is a board-level problem
A batch record is the single document that proves what you made, from what, on which line, to which spec, and who signed for it. On most high-production floors it still starts as handwriting on a clipboard, then gets re-keyed into a system hours or days later, if at all. That gap is not a floor-mechanics detail. It is where three separate costs originate, and each one lands on the executive who owns the P&L.
Cost 01 · Recall exposure
A trace you cannot run fast
When a lot is in question, the batch record is the evidence. If it lives in binders, a trace becomes phone calls and hours instead of a query. The scope of a recall is set by how precisely and quickly you can prove what a bad lot touched, and paper widens both.
Cost 02 · Audit and regulatory risk
Records you cannot produce on demand
Auditors and regulators increasingly expect records available on a clock and in a sortable electronic form. A handwritten batch sheet cannot reliably meet either. That is a compliance gap the owner answers for, not the operator.
Cost 03 · Margin and giveaway
Loss you only see at month end
Yield, scrap, and giveaway written on paper are invisible until someone tallies them later. By then the run is gone. A record that captures reconciliation per batch turns a month-end surprise into a same-shift correction.
Cost 04 · Re-keying and key-person risk
Labor spent re-typing what already exists
Every paper sheet is later typed into a system by a person. That is paid hours converting handwriting into data, plus the risk that the one person who reconciles it all is out for a week. It is the most legible line to cut.
So the executive question is not which quality software to buy. It is whether the record your whole compliance and margin story rests on is legible to a system at all. An AI copilot, a scheduling agent, or a predictive model cannot read a batch record that only exists as handwriting. The record has to be digital before any of the AI you are being pitched has something to stand on. That is why this belongs in Phase 1, not a later phase.
What a good batch record captures: the six sections
A complete batch record is not a longer form. It is six sections, each tied to one thing you have to be able to prove. Below are the section headings and what each one is for. The full field-by-field list under every section, the part your quality lead builds from, opens when you enter your work email.
Section 01
Identification
Proves: which batch this is
The batch or lot ID and everything that pins it to one run: product, spec version, line and equipment, date, shift, operator. Every reading downstream ties back here.
Section 02
Materials & inputs
Proves: what went into it
Every component consumed, with supplier lot, quantity, and a weigh-verify sign-off. This is the half of a trace that walks backward from the finished lot to incoming material.
Section 03
Process & in-process controls
Proves: what actually happened
Each process step with its setpoint, the actual reading, the time, and the equipment. The difference between what was planned and what the record shows is where quality problems hide.
Section 04
Quality checks & holds
Proves: it met spec
Every test against its limit, with result, disposition, deviation, and sign-off. A nonconformance is either contained inside this record or it walks out the door in a shipment.
Section 05
Yield & reconciliation
Proves: what it cost you
Theoretical versus actual yield, scrap, and reconciliation. This is where giveaway and loss become visible per batch instead of surfacing in a month-end variance nobody can explain.
Section 06
Review & release
Proves: who is accountable
Reviewer, quality release, electronic signature, and timestamp. The batch is disposed by a named person on a clock, so accountability is in the record rather than in someone's memory.
Want to know how much of your floor still runs on records like this before you build the template? The AI Readiness Checklist walks you through it, and the ROI Calculators & Tools put a dollar figure on the re-keying and giveaway using your own inputs.
The full template
Get the field-by-field template.
You have seen the six sections. Enter your work email and the full field list under each opens right here on this page, and a copy goes to your inbox to hand to your quality lead.
Every field under all six sections, with what it captures and why it matters
The fields that make a lot trace fast, called out where they sit
The one public rule that sets the clock on lot traces, with a link
Work email only. We use it to send the template and nothing else you did not ask for. Unsubscribe anytime.
Unlocked. The full template is open below, and a copy is on its way to your inbox. If you checked the box, a Harmony engineer will reach out to walk it onto one of your lines.
The full record, six sections, in the order work moves through the station. Fields marked trace are the ones a fast lot trace depends on.
Section 01 · Identification
Which batch this is
Filled once at the start of the run. Every later entry references this block, so it has to be unambiguous and captured before the first material moves.
Batch / Lot IDtrace. The unique identifier for this run. Everything else in the record, and every trace, hangs off this one field.
Product / SKUWhat was made, by the code your systems use, not a free-text name that drifts between shifts.
Specification / formula versionThe exact spec or recipe revision in force for this run, so a later change cannot be confused with what was actually built.
Line / cell / equipment IDtrace. Where it ran and on what asset, so a problem traced to a machine can be scoped to every lot that touched it.
Production date & time startedWhen the run opened, timestamped rather than written, so ordering and latency are exact.
Shift / crewWhich shift and crew ran it, so a pattern tied to a shift becomes visible across many batches.
Operator(s) of recordWho ran the batch, captured as an identity, not initials a stranger cannot resolve.
Section 02 · Materials & inputs
What went into it
One row per component consumed. This is the backward half of a trace, from the finished lot to every incoming material and its supplier.
Component / material IDThe item consumed, by code, matched to the bill of materials for the spec version above.
Supplier lot / receipt numbertrace. The incoming lot this material came from. Without it, a supplier-side problem cannot be scoped to your finished lots.
Quantity used (planned vs actual)What the spec called for and what was actually charged, so overage and substitution are visible.
Unit of measureThe unit the quantity is in, recorded explicitly so no reconciliation later has to guess it.
Weigh / dispense verificationThe measured weight or count at the point of charge, from the scale or reader where possible rather than eyeballed.
Verified byThe second identity that confirmed the material and quantity, so a mischarge has an accountable check.
Expiry / retest dateWhere the material carries one, so nothing out of date enters the batch unnoticed.
Section 03 · Process & in-process controls
What actually happened
One row per controlled step. The point of this section is to record the actual reading next to the target, not to reprint the target as if it were the outcome.
Process stepThe named step, in sequence, so the record reads as the run actually unfolded.
Parameter & setpoint / targetWhat was being controlled and to what target, for example temperature, pressure, speed, time, or fill weight.
Actual readingThe measured value, pulled from the equipment where it can be, so the record shows reality rather than intent.
Tolerance / acceptable rangeThe band the actual has to fall in, so an out-of-range reading is self-evident in the record.
Timestamptrace. When the step happened, so sequence and duration can be reconstructed and latency measured.
Equipment / instrument ID & calibrationWhich instrument produced the reading and whether it was in calibration, so a suspect reading can be run down to a device.
Performed / verified byWho executed and, where required, who verified the step.
Section 04 · Quality checks & holds
It met spec
One row per quality test or hold. A nonconformance is either contained inside this section with a disposition, or it leaves the plant in a shipment.
Test / check nameThe quality check performed, tied to the acceptance criteria for this product.
Specification limitThe pass criterion, recorded alongside the result so no one has to look it up to read the record.
Result & pass / failThe measured result and the plain call. A blank here is a hole in the release, not a minor omission.
Sample / retain referencetrace. The sample or retain this result came from, so a later question can be re-tested against the same material.
Deviation / nonconformance referenceThe linked deviation record where a result is out of spec, so the exception is tracked rather than erased.
DispositionAccept, rework, hold, or reject, recorded as an explicit decision on the batch.
Checked / approved byThe quality identity that signed the result, distinct from the operator who ran the step.
Section 05 · Yield & reconciliation
What it cost you
Closed at end of run. This is the section that turns giveaway and loss from a month-end variance into a same-shift number the P&L can act on.
Theoretical yieldWhat the batch should have produced, from the spec, as the baseline everything is measured against.
Actual yield / good unitsWhat it actually produced and passed, so the gap to theoretical is explicit on the record.
Scrap / rework / wasteWhat was lost and to what cause, so recurring loss on a product or line is countable rather than anecdotal.
Giveaway / overfillProduct given away above target, where applicable. Written on paper it is invisible; captured per batch it is a margin lever.
Reconciliation & varianceInputs consumed against outputs produced, with the variance and its explanation, so an unaccounted gap gets a reason attached.
Reconciled byThe identity that closed the numbers, so the reconciliation is owned rather than orphaned.
Section 06 · Review & release
Who is accountable
The disposition of the whole batch. This is what an auditor asks to see first, and what a recall investigation stands on.
Record complete / all fields enteredThe confirmation that no field above is blank, so an incomplete record cannot be released by default.
Reviewed by (production)The production sign-off that the run and its readings are as recorded.
Quality review & release decisionThe quality decision to release, reject, or hold the finished batch, as a named accountable act.
Exceptions / deviations closedConfirmation that every deviation raised in Section 04 is resolved before release, so nothing open ships.
Electronic signature & timestamptrace. The signed, time-stamped release, so accountability and timing are in the record and not reconstructed later.
Disposition to inventory / shipmentWhere the released batch went, closing the forward half of the trace from this lot onward.
Why the trace-marked fields matter, in one place
The fields marked trace are what let you answer a recall or an audit as a query instead of a fire drill. When those fields are digital and linked, walking from a finished lot back to every supplier lot and forward to every shipment is a search, not a week of pulling binders. What that trace is now expected to take is not a matter of opinion. Under the FDA Food Traceability Rule, which implements section 204 of the Food Safety Modernization Act, covered firms must make required traceability records available to an authorized FDA representative within 24 hours of a request, and during an outbreak, recall, or other public health threat must provide the required information in an electronic sortable spreadsheet within that same window. A paper batch record can reliably meet neither condition.
Source: 21 CFR 1.1455, paragraphs (c)(1) and (c)(3)(ii), FDA Food Traceability Rule.
Where this fits: Phase 1, the data foundation
Digitizing the batch record is not a quality-department project that happens to help. It is the opening move of the sequence every plant runs through, and it sits squarely in Phase 1. You get records off paper at the station, you connect the systems that already hold pieces of the picture, and you unify it into one live layer that is entered once. Only after that does the AI you are being pitched have something real to read.
Phase 1
Lay the Data Foundation · Digitization
Every pen-and-paper record, the batch record first among them, digitized at the station, every software system connected, and all of the data unified into one live layer.
Phase 2
Production & Operations Scale
With batch data live, operations turn proactive: real-time yield and quality, the AI scheduling board, predictive maintenance before failure.
Phase 3
AI-Native Operations
Agents read the live batch layer and act on it: quality signals, trace on demand, release copilots. Humans approve.
The board-level decision is whether to spend on AI at all before this foundation exists. The honest answer is that AI bought on top of paper records has nothing to stand on, which is why failed AI purchases so often trace back to missing or disconnected data rather than a weak model. Digitizing the batch record is the least glamorous and most fixable place to start, because the work is known rather than experimental.
Rather have this built onto one of your lines?
Putting this template onto a real line is the first week of a Harmony pilot: forward-deployed engineers on-site, digitizing the batch record at the station alongside your team. The pilot is a fixed $15,000 to $20,000 one time, runs 4 to 6 weeks, with working software in your plant by the end of the pilot. Phase 1 first, because that is the order it has to happen in. See what the live layer looks like.
